Mendelian Randomization for Nutrients and Migraine

ISEF Category: Translational Medical Science

Subcategory: Disease Prevention  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

Migraine can run in families, but family history does not always mean cause. Genetics can help you separate what is linked from what might actually change risk. That gives you a real research question, not just a guess. You can use public data to test whether certain nutrients may matter more than others.

What Is It?

This project uses Mendelian randomization, a method that treats genetic variants like natural markers of exposure. If some gene variants are linked to higher vitamin D, omega-3, or magnesium levels, you can test whether those variants also line up with lower migraine frequency. Think of it like using a coin that was stamped at birth. You did not choose the stamp, so it can help you ask whether the nutrient itself may play a causal role.

The big idea is simple. A normal correlation can confuse cause and effect. Maybe people with migraines eat differently because of the migraine, or maybe something else affects both diet and pain. Mendelian randomization tries to sidestep that problem by using inherited differences as the starting point. If the nutrient still predicts migraine after that test, you have stronger evidence that the nutrient could matter.

Why This Is a Good Topic

This is a strong science fair topic because the data are public, the question is narrow, and the analysis has a clear yes-or-no shape. You can compare more than one nutrient, rank their likely impact, and connect the result to a real health problem that affects millions of people. You will also learn a real research skill, reading GWAS summary stats, checking instrument strength, and thinking about causality instead of just correlation.

Research Questions

• How does genetically predicted vitamin D level affect migraine frequency in public GWAS data?
• How does genetically predicted omega-3 level affect migraine frequency in public GWAS data?
• How does genetically predicted magnesium level affect migraine frequency in public GWAS data?
• To what extent do the three nutrient signals differ in estimated causal effect on migraine frequency?
• Which nutrient has the strongest Mendelian-randomization evidence for lowering migraine frequency?
• Does the result stay similar when you test alternative GWAS datasets for migraine?
• What is the effect of removing weak genetic instruments on the nutrient-migraine estimate?

Basic Materials

• Laptop with internet access.
• PubMed and Google Scholar access through a browser.
• GWAS summary statistics for migraine and nutrient-related traits.
• Spreadsheet software such as Google Sheets or Excel.
• R or Python installed on your computer.
• Basic statistics reference on odds ratios, confidence intervals, and p values.
• Notebook for tracking dataset sources and analysis choices.

Advanced Materials

• Laptop or desktop with enough memory for summary-statistic analysis.
• R with TwoSampleMR or similar Mendelian-randomization package.
• Python with pandas, numpy, scipy, and matplotlib.
• Access to multiple GWAS summary-statistics datasets for replication.
• LD reference panel for clumping instruments.
• MR sensitivity-analysis tools for pleiotropy checks.
• Version control software such as Git for tracking code changes.

Software & Tools

• R: Runs Mendelian-randomization workflows and sensitivity tests on summary GWAS data.
• TwoSampleMR: Helps you select instruments, estimate causal effects, and run common MR checks.
• Python: Cleans summary statistics, merges datasets, and makes plots for your results.
• PLINK: Clumps linked genetic variants so your instruments stay independent.
• Google Sheets: Organizes study notes, dataset links, and result tables if you are starting out.

Experiment Steps

1. Define the single migraine outcome you will study, and decide whether you want frequency, diagnosis, or severity as your endpoint.
2. Pick one nutrient at a time, and identify GWAS summary stats that measure its genetic proxy clearly.
3. Choose genetic instruments, then plan how you will filter weak or highly linked variants.
4. Match the exposure and outcome datasets, and decide how you will handle ancestry overlap, sample overlap, and missing SNPs.
5. Build a main analysis plan, then add sensitivity tests that check whether one bad variant is driving the result.
6. Rank the nutrients by effect size, confidence interval, and consistency, then choose the strongest one for a follow-up self-experiment idea.

Common Pitfalls

• Using a nutrient GWAS that measures a proxy trait, which can weaken the causal interpretation.
• Mixing migraine frequency with migraine diagnosis, which changes the meaning of the outcome.
• Keeping weak genetic instruments, which can make the MR estimate unstable.
• Ignoring pleiotropy, which happens when one variant affects migraine through another pathway.
• Comparing results from different datasets without checking ancestry, sample overlap, or phenotype definitions.

What Makes This Competitive

A stronger version of this project does more than report one MR estimate. You compare several nutrients, run sensitivity analyses, and explain why one signal looks more trustworthy than the others. You can also test whether the result holds across more than one migraine dataset. A thoughtful ranking system for follow-up self-experimentation makes the project feel more like real translational research.

Project Variations

• Use migraine diagnosis instead of migraine frequency as the outcome, then compare whether the causal signals change.
• Swap in dietary intake GWAS for biomarker GWAS, then test whether the proxy choice changes the result.
• Add a second-stage analysis that compares self-reported supplement use with the MR ranking, then look for agreement or conflict.

Learn More

• NIH Office of Dietary Supplements Fact Sheets: Background on vitamin D, omega-3, and magnesium, found by searching the NIH site.
• PubMed: Search for review articles on migraine, nutrient status, and Mendelian randomization to see how researchers frame the question.
• GWAS Catalog: Find public genome-wide association study summary data and trait definitions for exposure and outcome planning.
• IEU OpenGWAS: Access public GWAS summary statistics and common MR-friendly datasets through the University of Bristol resource.
• Nature Reviews Methods Primers: Search for Mendelian randomization primers to learn the core assumptions and common bias sources.